降水集合预报集成方法研究

Study on Precipitation Multimodel Super Ensemble Forecast Technique

基于TIGGE(the THORPEX Interactive Grand Global Ensemble)资料,对中国气象局(CMA)、欧洲中期天气预报中心(ECMWF)、美国国家环境预报中心(NCEP)和日本气象厅(JMA)的集合数值预报结果进行降水集成。采用算术平均法、TS评分集成法和BS评分集成法在我国东南地区进行降水集成,对比分析结果表明:基于TS评分的多模式降水集成无论在分区降水评分中,还是在东南地区的台风型降水和非台风型降水实例中,都有效地改进了大雨以上的降水预报效果;基于BS评分的集成方法和算数平均集成法预报效果次之。东南地区5个子区域的降水集成试验结果表明:各子区域基于TS评分集成后降水的平均绝对误差普遍小于基于BS评分后的降水平均绝对误差。广东东南和浙江北部区域基于TS集成后的降水TS评分值最优,浙闽沿海和广东西北部区域基于TS集成后的降水TS评分次之,处于中上水平。基于算术平均集成和BS集成的降水的TS评分值只有在广东东南区域表现出较好的效果。

Ensemble-weight methods based on the Threat Score （TS）, the Brier Score （BS） and the Arith- metic Average （AA） were employed to combine the precipitation prediction of four ensemble numerical prediction systems （CMA, ECMWF, NCEP and JMA） from the TIGGE over the southeast of China and its five sub regions. The results of comparison analysis of the four ensemble numerical prediction systems indicated that the super ensemble based on TS-weight always provides more skillful performance in any of the four ensemble numerical prediction systems regionally and shows the best prediction performance of rainfall impacted by typhoon landing or not. The grand ensemble based on BS weight and AA-weight also shows the better skillful performance. Moreover, verification of super ensemble method in five sub-regions showed that the super ensemble precipitation based on TS-weight always has a less mean absolute error （MAE） and higher TS score in most sub-regions than that based on BS-weight and AA-weight. The super ensemble precipitation based on TS-weight provides the best performance in the southeast of Guangdong and northern Zhejiang and the secondary performance along the coast of Zhejiang and Fujian and the north west of Guangdong. The skillful performance based on BS weight and AA-weight is provided only in thesoutheast of Guangdong.